Asset Pricing Restrictions on Predictability: Frictions Matter

Abstract

U.S. stock portfolios sorted on size, momentum, transaction costs, M/B, I/A and ROA ratios, and industry classification show considerable levels and variation of return predictability, inconsistent with asset pricing models. This means that a predictable risk premium is not equal to compensation for systematic risk as implied by asset pricing theory (Kirby 1998). We show that introducing market frictions relaxes these asset pricing moments from a strict equality to a range. Empirically, it is not short sales constraints but transaction costs (below 35 basis points) that help to reconcile the observed predictability with the Fama-French-Carhart four-factor model and the Chen-Novy-Marx-Zhang three factor model, and partly with the Durable Consumption model. Across the sorts, predictability in industry returns can be reconciled with all models considered with only 25 basis points transaction costs, whereas for momentum and ROA portfolios up to 115 basis points are needed.

Keywords: time-series predictability, cross-sectional predictability, asset pricing tests, market frictions

JEL Classification: G12, G17